Variance-Optimized Rao-Blackwellization for Mitigating Error Propagation in a Decision Feedback Equalization

A variance-optimized Rao-Blackwellization (VO-RB) soft decision scheme is presented for mitigating error propagation in a decision feedback equalizer (DFE). The VO-RB soft decision scheme uses optimized variance for the RB decision block to minimize a cost function, instead of the tentative equalizer output variance as in the original Rao-Blackwellization approach. This VO-RB decision device can be employed stand-alone or jointly with any other soft scheme for the joint optimization of DFE filter to minimize bit error rate (BER) or other cost function. In this paper, the VO-RB is employed by itself and with scaling soft scheme for HDTV and EDGE systems, and simulation results show the performance gain of the VO-RB with/ without scaled feedback regressor over other conventional decision devices.